AAAI 2025 Papers — Page 10
AAAI Conference on Artificial Intelligence · 3028 papers
Efficient Rectification of Neuro-Symbolic Reasoning Inconsistencies by Abductive Reflection
Wen-Chao Hu (Nanjing University), Zhi-Hua Zhou (Nanjing University)
OptimizationComputational EfficiencyConvolutional Neural NetworkGraph Neural NetworkReinforcement LearningImageText
🎯 What it does: A new neural symbolic reasoning framework called Abductive Reflection (ABL-Refl) is proposed, which quickly detects predictions inconsistent with domain knowledge by parallelly generating reflection vectors during neural network output, and corrects errors using abduction.
Efficient Reinforcement Learning in Probabilistic Reward Machines
Xiaofeng Lin (Boston University), Xuezhou Zhang (Boston University)
Reinforcement Learning
🎯 What it does: This paper studies reinforcement learning in Markov decision processes with probabilistic reward machines (PRM), proposing the UCBVI-PRM algorithm and providing an approximate optimal regret upper bound.
Efficient Reinforcement Learning Through Adaptively Pretrained Visual Encoder
Yuhan Zhang (Institute of Automation, Chinese Academy of Sciences), Shan Yu (Institute of Automation, Chinese Academy of Sciences)
Reinforcement LearningContrastive LearningImage
🎯 What it does: This paper proposes the APE framework, which significantly enhances the generalization and sampling efficiency of visual reinforcement learning by pre-training the visual encoder with adaptive data augmentation during the pre-training phase and freezing the first few layers of the encoder while only fine-tuning the last layer during the policy learning phase.
Efficient Robustness Evaluation via Constraint Relaxation
Chao Pan (Southern University of Science and Technology), Xin Yao (Lingnan University)
OptimizationComputational EfficiencyAdversarial AttackConvolutional Neural NetworkImage
🎯 What it does: Proposes an attack method to accelerate model robustness evaluation through dynamic relaxation of perturbation constraints (Constraint Relaxation Attack, CR Attack).
Efficient Self-Supervised Video Hashing with Selective State Spaces
Jinpeng Wang (Tsinghua University), Shu-Tao Xia (Tsinghua University)
RetrievalComputational EfficiencyRecurrent Neural NetworkContrastive LearningVideo
🎯 What it does: A self-supervised video hashing method S5VH based on the Mamba state space model is proposed, utilizing bidirectional Mamba layers and a self-local-global learning strategy to achieve efficient temporal modeling and hash code generation.
Efficient Traffic Prediction Through Spatio-Temporal Distillation
Qianru Zhang (University of Hong Kong), Hongzhi Yin (University of Queensland)
Autonomous DrivingComputational EfficiencyKnowledge DistillationGraph Neural NetworkTime Series
🎯 What it does: This paper proposes a lightweight traffic flow prediction framework called LightST, which significantly improves inference speed while maintaining high accuracy by distilling spatial and temporal knowledge from a high-order graph neural network teacher model to an MLP student model.
Efficient Training of Neural Fractional-Order Differential Equation via Adjoint Backpropagation
Qiyu Kang (University of Science and Technology of China), Wee Peng Tay (Nanyang Technological University)
ClassificationOptimizationComputational EfficiencyGraph Neural NetworkGraphOrdinary Differential Equation
🎯 What it does: An adaptive backpropagation method based on reverse solving enhanced fractional differential equations (FDE) is proposed, significantly reducing the memory consumption during the training process of neural FDE while maintaining the same performance as traditional forward automatic differentiation methods.
Efficiently Enhancing Long-term Series Forecasting via Ultra-long Lookback Windows
Suxin Tong (Wuhan University of Technology), Jingling Yuan (Wuhan University of Technology)
OptimizationComputational EfficiencyTransformerTime Series
🎯 What it does: The IRPA framework is proposed and implemented, which extracts key information from an ultra-long lookback window through the Input Refinement Module (IRM) and the Prediction Assistance Module (PAM), and uses this information to enhance the accuracy of long-period time series forecasting.
EfficientVMamba: Atrous Selective Scan for Light Weight Visual Mamba
Xiaohuan Pei (University of Sydney), Chang Xu (University of Sydney)
ClassificationObject DetectionSegmentationConvolutional Neural NetworkImage
🎯 What it does: A lightweight visual model called EfficientVMamba is proposed, which combines state space models (SSM) with convolution, and achieves efficient extraction of global and local features through sparse scanning (ES2D) and dual-channel fusion.
EFSkip: A New Error Feedback with Linear Speedup for Compressed Federated Learning with Arbitrary Data Heterogeneity
Hongyan Bao (Singapore Management University), Zhize Li (Pennsylvania State University)
OptimizationFederated LearningComputational Efficiency
🎯 What it does: This paper proposes the EFSkip framework for compressing communication in federated learning, improving upon traditional error feedback (EF) methods, allowing for arbitrary data heterogeneity and achieving linear acceleration.
EGSRAL:An Enhanced 3D Gaussian Splatting Based Renderer with Automated Labeling for Large-Scale Driving Scene
Yixiong Huo (Advanced Micro Devices), Emad Barsoum (Advanced Micro Devices)
Object DetectionAutonomous DrivingGaussian SplattingImage
🎯 What it does: An enhanced 3D Gaussian scattering rendering framework named EGSRAL is proposed, which can achieve new perspective rendering in large driving scenes using only image data and automatically generate corresponding 2D/3D annotations.
EigenSR: Eigenimage-Bridged Pre-Trained RGB Learners for Single Hyperspectral Image Super-Resolution
Xi Su (Chongqing University), Xichuan Zhou (Chongqing University)
RestorationSuper ResolutionTransformerSupervised Fine-TuningImage
🎯 What it does: This study addresses the single hyperspectral image super-resolution (single-HSI-SR) problem and proposes a new framework called EigenSR. This framework first transfers a pre-trained RGB Transformer (IPT) to the feature map (eigenimage) domain for single-channel detail learning, and then utilizes Iterative Spectral Regularization (ISR) to restore spectral consistency during inference, thereby enhancing SR performance in both spatial and spectral dimensions simultaneously.
ELDER: Enhancing Lifelong Model Editing with Mixture-of-LoRA
Jiaang Li (University of Science and Technology of China), Zhendong Mao (University of Science and Technology of China)
TransformerLarge Language ModelMixture of ExpertsText
🎯 What it does: A lifelong model editing framework named ELDER is proposed, utilizing a hybrid LoRA combined with routers to achieve continuous adapter allocation, thereby maintaining knowledge consistency and improving robustness during multiple edits.
Elevating Flow-Guided Video Inpainting with Reference Generation
Suhwan Cho (Yonsei University), Joon-Young Lee (Adobe Research)
RestorationGenerationDiffusion modelOptical FlowVideoBenchmark
🎯 What it does: A video restoration framework RGVI is proposed, which combines optical flow-guided dual pixel propagation and large-scale diffusion models, enabling high-quality object removal and content generation.
Eliciting Causal Abilities in Large Language Models for Reasoning Tasks
Yajing Wang (BNU-HKBU United International College), Bo Han (Hong Kong Baptist University)
OptimizationExplainability and InterpretabilityMeta LearningTransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: By treating prompts as processing variables, this study uses causal inference to estimate their impact on the accuracy of reasoning results from large language models, and based on this, generates better prompts to enhance the model's performance on reasoning tasks.
Eliminating Majority Illusion Is Easy
Jack Dippel (McGill University), Adrian Vetta (McGill University)
Graph
🎯 What it does: Completely eliminating the majority illusion in social networks (i.e., ensuring that blue nodes occupy at least half of each node's neighborhood) and proving that this goal can be achieved in polynomial time.
ELLA-V: Stable Neural Codec Language Modeling with Alignment-Guided Sequence Reordering
Yakun Song (Shanghai Jiao Tong University), Xie Chen (Shanghai Jiao Tong University)
GenerationTransformerTextAudio
🎯 What it does: A zero-shot text-to-speech generation framework based on a language model, ELLA-V, is proposed, utilizing an inserted phoneme-acoustic interleaved sequence for fine-grained phoneme-level control.
Embedding Robust Watermarking into Pattern to Protect the Copyright of Ceramic Artifacts
Lei Tan (Fudan University), Chunlei Bao (Fudan University)
Convolutional Neural NetworkImage
🎯 What it does: Embedding watermarks through invisible templates in ceramic processes to protect copyright.
EMControl: Adding Conditional Control to Text-to-Image Diffusion Models via Expectation-Maximization
He Wang (Nanjing University of Science and Technology), Jinhui Tang (Nanjing University of Science and Technology)
SegmentationGenerationData SynthesisPose EstimationDepth EstimationDiffusion modelImage
🎯 What it does: This paper proposes EMControl, a training-free method that learns a latent space forward operator in text-image diffusion models and uses the Expectation-Maximization (EM) framework during the sampling phase, while estimating the conditional correction term ϕ and the correction term ψ, thereby achieving control over various conditions (such as edges, depth, segmentation, pose, style, etc.).
Emergence-Inspired Multi-Granularity Causal Learning
Hanwen Luo (Shandong University), Qingzhong Li (Shandong University)
OptimizationExplainability and InterpretabilityGraph Neural NetworkAuto EncoderTabular
🎯 What it does: A multi-granularity causal learning method called EMCausal based on 'emergent phenomena' is proposed, which can simultaneously mine the causal structure between micro variables and macro aggregate variables.
EMHI: A Multimodal Egocentric Human Motion Dataset with HMD and Body-Worn IMUs
Zhen Fan (ByteDance), Yang Zhang (ByteDance)
Pose EstimationRecurrent Neural NetworkMultimodalityTime SeriesBenchmark
🎯 What it does: Proposed the EMHI multimodal perspective human motion dataset and the MEPoser benchmark method, combining perspective images and IMU to achieve VR HPE.
EmoReg: Directional Latent Vector Modeling for Emotional Intensity Regularization in Diffusion-based Voice Conversion
Ashishkumar Prabhakar Gudmalwar (Sony Research India), Rajiv Ratn Shah (Indraprastha Institute of Information Technology)
GenerationData SynthesisDiffusion modelStochastic Differential EquationAudio
🎯 What it does: This paper proposes an emotion voice conversion framework called EmoReg based on diffusion models, which can finely adjust the target emotional intensity while maintaining semantic content.
EMPLACE: Self-Supervised Urban Scene Change Detection
Tim Alpherts (University of Amsterdam), Nanne van Noord (University of Amsterdam)
SegmentationAnomaly DetectionTransformerContrastive LearningImage
🎯 What it does: A large-scale urban street scene change detection dataset AC‑1M has been constructed, and a self-supervised method EMPLACE has been proposed to learn urban change features.
Empowering LLMs with Pseudo-Untrimmed Videos for Audio-Visual Temporal Understanding
Yunlong Tang (University of Rochester), Chenliang Xu (University of Rochester)
RecognitionGenerationData SynthesisTransformerLarge Language ModelSupervised Fine-TuningVideoMultimodalityAudio
🎯 What it does: Designed and constructed the PU-VALOR pseudo-uncut audio-video dataset, using this dataset to train the AVicuna model to achieve precise alignment and localization of audio-video temporal events.
Empowering Self-Learning of LLMs: Inner Knowledge Explicitation as a Catalyst
Shijue Huang (Harbin Institute of Technology), Ruifeng Xu (Harbin Institute of Technology)
TransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: The SKE-Learn framework is proposed, which enables LLM to explicitly extract intrinsic knowledge, verify it, and utilize this knowledge for reasoning, thereby achieving reliable self-learning data filtering.
Enabling Region-Specific Control via Lassos in Point-Based Colorization
Sanghyeon Lee (Korea Advanced Institute of Science and Technology), Jaegul Choo (Korea Advanced Institute of Science and Technology)
Image TranslationData SynthesisTransformerImage
🎯 What it does: A lasso tool is proposed for controlling the color propagation range in point-based interactive image coloring to address the color collapse problem.
Encoder of Thoughts: Enhancing Planning Ability in Language Agents Through Structural Embedding
Yuxiang Zhang (Beijing Jiaotong University), Jitao Sang (Beijing Jiaotong University)
Graph Neural NetworkLarge Language ModelTextGraph
🎯 What it does: Proposes the Encoder of Thoughts (EoT) framework, which encodes planning graphs into structural embeddings through graph neural networks and injects them into LLMs to enhance the planning and reasoning capabilities of language agents.
ENCODER: Entity Mining and Modification Relation Binding for Composed Image Retrieval
Zixu Li (Shandong University), Weili Guan (Harbin Institute of Technology)
RetrievalTransformerVision Language ModelContrastive LearningImageMultimodality
🎯 What it does: A new CIR model called ENCODER is proposed, which can mine the correspondence between visual entities and modification actions from multimodal queries and bind them.
End-to-End Autonomous Driving Through V2X Cooperation
Haibao Yu (University of Hong Kong), Zaiqing Nie (AIR Tsinghua University)
Autonomous DrivingTransformerPoint Cloud
🎯 What it does: Designed and implemented an end-to-end V2X collaborative autonomous driving framework called UniV2X, which integrates perception, online mapping, occupancy prediction, and planning modules into a single network, and achieves cross-view fusion through sparse-dense mixed data transmission.
END^2: Robust Dual-Decoder Watermarking Framework Against Non-Differentiable Distortions
Nan Sun (Huazhong University of Science and Technology), Hefei Ling (Huazhong University of Science and Technology)
Data SynthesisKnowledge DistillationConvolutional Neural NetworkImage
🎯 What it does: Proposed the END2 dual decoder framework to address the robustness issue of deep learning image watermarking against non-differentiable perturbations.
Energy vs. Noise: Towards Robust Temporal Action Localization in Open-World
Chenyu Mu (Xidian University), Cheng Deng (Xidian University)
OptimizationMeta LearningVideo
🎯 What it does: This paper proposes the Energy-Driven Meta Purifier (EDMP), which utilizes an energy-driven meta-learning framework to remove boundary and category noise in Temporal Action Localization (TAL), enhancing the model's robustness against open-world noise.
Energy-Guided Optimization for Personalized Image Editing with Pretrained Text-to-Image Diffusion Models
Rui Jiang (Zhejiang University), Xi Li (Zhejiang University)
GenerationOptimizationDiffusion modelImageStochastic Differential Equation
🎯 What it does: This paper proposes a training-free and inversion-free energy-guided optimization framework for personalized image editing, which gradually optimizes the latent code of the target image under the guidance of text and image energy, achieving cross-category object replacement.
Enhance Vision-Language Alignment with Noise
Sida Huang (Northwestern Polytechnical University), Xuelong Li (China Telecom)
ClassificationDomain AdaptationConvolutional Neural NetworkContrastive LearningImage
🎯 What it does: This paper proposes a CLIP fine-tuning method based on positive incentive noise (PiNI), which enhances visual-language alignment by injecting learned noise into the visual and text encoders, thereby achieving better performance on downstream tasks.
Enhanced Denesity Peak Clustering for High-Dimensional Data
Zhongli Wang (Hangzhou Normal University), Weiguo Sheng (University of Technology Sydney)
OptimizationSupervised Fine-TuningTabularBiomedical Data
🎯 What it does: An Enhanced Density Peak Clustering (EDPC) method is proposed, which combines dimensionality reduction using a multilayer perceptron and hierarchical label assignment, significantly improving clustering performance on high-dimensional data.
Enhanced Importance Sampling Through Latent Space Exploration in Normalizing Flows
Liam Anthony Kruse (Stanford University), Mykel J. Kochenderfer (Stanford University)
Autonomous DrivingOptimizationComputational EfficiencyFlow-based ModelSequential
🎯 What it does: This study proposes performing importance sampling in the normalized flow latent space to improve the efficiency of rare event simulation.
Enhanced Sample Selection with Confidence Tracking: Identifying Correctly Labeled Yet Hard-to-Learn Samples in Noisy Data
Weiran Pan (Huazhong University of Science and Technology), Yong Deng (State Grid Fujian Electric Power Company)
ClassificationConvolutional Neural NetworkImage
🎯 What it does: A sample selection method based on confidence tracking is proposed, which can identify correct but hard-to-learn samples in image classification tasks with noisy labels.
Enhancing Adversarial Transferability with Adversarial Weight Tuning
Jiahao Chen (Zhejiang University), Shouling Ji (Zhejiang University)
Adversarial AttackConvolutional Neural NetworkTransformerImage
🎯 What it does: This paper studies the intrinsic mechanisms of transferable adversarial attacks and proposes the Adversarial Weight Tuning (AWT) method, which adjusts the parameters of a surrogate model under no data conditions to enhance the transferability of adversarial samples across models.
Enhancing Audiovisual Speech Recognition Through Bifocal Preference Optimization
Yihan Wu (Carnegie Mellon University), Shinji Watanabe (Carnegie Mellon University)
RecognitionOptimizationTransformerLarge Language ModelSupervised Fine-TuningVideoMultimodalityAudio
🎯 What it does: Train the AV-ASR model through dual-focus preference optimization (input side and output side) to improve speech recognition accuracy in real-world videos.
Enhancing Chain of Thought Prompting in Large Language Models via Reasoning Patterns
Yufeng Zhang (Institute of Automation, Chinese Academy of Sciences), Jinqiao Wang (Institute of Automation, Chinese Academy of Sciences)
Explainability and InterpretabilityTransformerLarge Language ModelPrompt EngineeringTextBenchmarkChain-of-Thought
🎯 What it does: A chain-of-thought demonstration selection method based on reasoning patterns, Pattern-CoT, is proposed to enhance the reasoning performance of large language models.
Enhancing Close-up Novel View Synthesis via Pseudo-labeling
Jiatong Xia (Australian Institute for Machine Learning), Lingqiao Liu (Australian Institute for Machine Learning)
RestorationGenerationData SynthesisNeural Radiance FieldGaussian SplattingImage
🎯 What it does: This paper improves the quality of near-field synthesis from viewpoints far from the training perspective through a pseudo-labeling learning strategy, particularly focusing on detail reconstruction at close distances.
Enhancing Contrastive Learning Inspired by the Philosophy of “The Blind Men and the Elephant”
Yudong Zhang (Tsinghua University), Yu Wang (Tencent)
ClassificationObject DetectionSegmentationRepresentation LearningContrastive LearningImage
🎯 What it does: Two methods, JointCrop and JointBlur, are proposed to control the augmentation parameters of positive sample pairs (the area ratio of Crop and the degree of Gaussian Blur) using joint distribution, generating more challenging positive sample pairs, and unifying them into a JointAugmentation framework.
Enhancing Decision-Making for LLM Agents via Step-Level Q-Value Models
Yuanzhao Zhai (National University of Defense Technology), Huaimin Wang (National University of Defense Technology)
OptimizationReinforcement Learning from Human FeedbackTransformerLarge Language ModelReinforcement LearningAgentic AITextBenchmark
🎯 What it does: Proposes the use of a step-level Q-value model collected based on MCTS to guide action selection in the multi-step decision-making process of LLM agents, improving decision quality.
Enhancing Diffusion Model with Auxiliary Information Mining-Exploration and Efficient Sampling Mechanism for Sequential Recommendation
Te Song (China University of Petroleum), Wanchun Dou (Nanjing University)
Recommendation SystemGraph Neural NetworkDiffusion modelAuto EncoderSequential
🎯 What it does: A sequence recommendation framework DAE4Rec based on diffusion models is proposed, which utilizes a graph autoencoder to mine auxiliary information to generate better conditional vectors, and employs non-Markovian 'skip-step' sampling and a compensator to accelerate inference.
Enhancing Elusive Clues in Knowledge Learning by Contrasting Attention of Language Models
Jian Gao (Tsinghua University), Ji Wu (Tsinghua University)
TransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: By comparing the attention weights of large models and small models, important clues that are difficult to capture in the text are identified, and data augmentation based on token-dropout is conducted to enhance knowledge learning efficiency.
Enhancing Entertainment Translation for Indian Languages Using Adaptive Context, Style and LLMs
Pratik Rakesh Singh (Sony Research India), Pankaj Wasnik (Sony Research India)
GenerationRetrievalDomain AdaptationTransformerLarge Language ModelPrompt EngineeringTextRetrieval-Augmented Generation
🎯 What it does: We propose the CASAT framework, which utilizes context retrieval and style extraction modules to generate dynamic prompts suitable for entertainment translation for large language models, thereby achieving cultural adaptation and emotional retention in automated dialogues.
Enhancing Fine-Grained Vision-Language Pretraining with Negative Augmented Samples
Yeyuan Wang (Northwestern Polytechnical University), Xiaoyan Cai (Northwestern Polytechnical University)
RetrievalRepresentation LearningTransformerContrastive LearningImageTextMultimodality
🎯 What it does: This paper proposes a joint VLP method for generating hard negative samples, utilizing a visual dictionary (VD) for fine-grained alignment of visual and textual features, and generating token-level negative samples in the visual modality through Negative Visual Augmentation (NVA), further enhancing the model's fine-grained visual-language understanding capability.
Enhancing Generalizability in Molecular Conformation Generation with METRIZATION-Informed Geometric Diffusion Pretraining
Xiaozhuang Song (Chinese University of Hong Kong), Tianshu Yu (Chinese University of Hong Kong)
GenerationDrug DiscoveryGraph Neural NetworkDiffusion modelGraph
🎯 What it does: Using distance geometry constraints (METRIZATION) for pre-training diffusion generative models and fine-tuning on real data, we propose the Metrization-Informed Geometric Diffusion (MIGDIFF) framework to enhance the generalization performance of molecular conformation generation.
Enhancing Generalizability via Utilization of Unlabeled Data for Occupancy Perception
Ruihang Li (Zhejiang University), Zhijie Pan (Zhejiang University)
Domain AdaptationAutonomous DrivingPoint Cloud
🎯 What it does: The UGOCC method is proposed, which utilizes unlabeled target domain data to enhance the cross-domain generalization performance of 3D occupancy perception models through semantic query adversarial fusion, self-supervised deep enhancement, and pseudo-label selection.
Enhancing Generalized Few-Shot Semantic Segmentation via Effective Knowledge Transfer
Xinyue Chen (King's College London), Sophia Tsoka (King's College London)
SegmentationConvolutional Neural NetworkImage
🎯 What it does: A generalized few-shot semantic segmentation method GFSS-EKT based on effective knowledge transfer is proposed, which achieves the transfer of base class knowledge to new classes through three main modules: prototype modulation, classifier calibration, and context consistency learning.
Enhancing Healthcare Recommendations: A Privacy-Protective and Interpretable Cross-Domain Framework
Xun Liang (Renmin University of China), Hongxun Jiang (Renmin University of China)
Recommendation SystemSafty and PrivacyExplainability and InterpretabilityGraph Neural NetworkLarge Language ModelVideoTextMultimodality
🎯 What it does: A cross-domain recommendation framework (HCR) for healthcare services that is interpretable and privacy-preserving has been constructed.
Enhancing Identity-Deformation Disentanglement in StyleGAN for One-Shot Face Video Re-Enactment
Qing Chang (Zhejiang University), Kun Zhou (Zhejiang University)
GenerationData SynthesisTransformerGenerative Adversarial NetworkVideo
🎯 What it does: One-shot facial video reenactment based on StyleGAN2 is implemented, proposing a two-stage model. The first stage decouples identity and expression/pose in the latent space through an identity encoder, a deformation encoder, and conditional fusion. The second stage uses a refinement network in the feature space to complete high-frequency details.
Enhancing Implicit Neural Representations via Symmetric Power Transformation
Weixiang Zhang (Tsinghua University), Zhi Wang (Tsinghua University)
RestorationSuper ResolutionNeural Radiance FieldImageVideoMultimodalityAudio
🎯 What it does: This paper proposes a symmetric power transformation for the input data of implicit neural representations (INR) to enhance their expressive capability and accelerate training;
Enhancing Large Language Model Performance with Gradient-Based Parameter Selection
Haoling Li (Tsinghua University), Peng Cheng (Microsoft Research)
TransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: This paper proposes Gradient-Mask Tuning (GMT), which selects parameters to update by masking small gradients based on gradient magnitude during the fine-tuning process, thereby reducing redundant updates and improving the performance of LLMs on multiple tasks.
Enhancing LLMs via High-Knowledge Data Selection
Feiyu Duan (Beihang University), Xunliang Cai (Meituan)
TransformerLarge Language ModelText
🎯 What it does: A multi-domain knowledge element pool is constructed, and a gradient-free high knowledge scorer (HKS) is proposed to select high-knowledge texts based on knowledge density and coverage for large-scale language model pre-training.
Enhancing Long-and Short-Term Representations for Next POI Recommendations via Frequency and Hierarchical Contrastive Learning
Jiajie Chen (Soochow University), Zhixu Li (Renmin University of China)
Recommendation SystemTransformerContrastive LearningSequential
🎯 What it does: This paper proposes an FHCRec model that jointly models long-term and short-term user interests through frequency domain enhanced Transformer and hierarchical contrastive learning, thereby improving the accuracy of next location recommendations.
Enhancing Low-Light Images: A Synthetic Data Perspective on Practical and Generalizable Solutions
Yu Long (Beijing Institute of Technology), Yuming Fang (Jiangxi University of Finance and Economics)
RestorationData SynthesisImage
🎯 What it does: A low-light image synthesis pipeline from RAW inverse ISP to sRGB is proposed, which can automatically generate an unlimited amount of aligned low-light-normal light paired data.
Enhancing Low-Rank Adaptation with Recoverability-Based Reinforcement Pruning for Object Counting
Haojie Guo (Northwestern Polytechnical University), Yuan Yuan (Northwestern Polytechnical University)
Object DetectionReinforcement LearningImage
🎯 What it does: This paper proposes a recoverable pruning method based on reinforcement learning, E3RP, which utilizes the pre-trained weights of the large model SAM for parameter pruning and low-rank adaptation in object counting tasks.
Enhancing Masked Time-Series Modeling via Dropping Patches
Tianyu Qiu (Fudan University), Xiaofeng Gao (Shanghai Jiao Tong University)
Anomaly DetectionRepresentation LearningTransformerTime Series
🎯 What it does: Introducing a random dropout subsequence (patch) strategy (DropPatch) in self-supervised pre-training of time series, where some patches are dropped before performing masked reconstruction, improving the model's learning quality;
Enhancing Multi-Hop Fact Verification with Structured Knowledge-Augmented Large Language Models
Han Cao (Institute of Information Engineering, Chinese Academy of Sciences), Songlin Hu (Institute of Information Engineering, Chinese Academy of Sciences)
Graph Neural NetworkTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringTextGraph
🎯 What it does: A structured knowledge-enhanced LLM network (LLM-SKAN) is proposed for multi-hop fact verification, where LLM first extracts fine-grained entity relationships and then a graph neural network integrates reasoning.
Enhancing Multi-Robot Semantic Navigation Through Multimodal Chain-of-Thought Score Collaboration
Zhixuan Shen (Southwest Jiaotong University), Tianrui Li (Southwest Jiaotong University)
Object DetectionRobotic IntelligenceConvolutional Neural NetworkVision Language ModelMultimodalityChain-of-Thought
🎯 What it does: A multi-robot semantic navigation framework MCoCoNav based on multimodal chain reasoning is proposed, which evaluates exploration value using a visual language model and achieves low-cost communication through a global semantic graph.
Enhancing Multi-View Classification Reliability with Adaptive Rejection
Wei Liu (Tongji University), Xiaodong Yue (Shanghai University)
ClassificationAnomaly DetectionContrastive LearningBiomedical DataMagnetic Resonance Imaging
🎯 What it does: A framework is proposed that introduces an adaptive rejection mechanism in multi-view classification, utilizing evidence theory to reject low-quality views and fuse trustworthy information.
Enhancing Multimodal Affective Analysis with Learned Live Comment Features
Zhaoyuan Deng (Columbia University), Kathleen McKeown (Columbia University)
ClassificationData SynthesisRepresentation LearningTransformerContrastive LearningVideoTextMultimodalityAudio
🎯 What it does: A large-scale multilingual multimodal live comment dataset LCAffect was constructed, and a video encoder was trained using contrastive learning to generate synthetic live comment features for videos without comments, aimed at enhancing multimodal sentiment analysis tasks.
Enhancing Multimodal Large Language Models Complex Reason via Similarity Computation
Xiaofeng Zhang (Shanghai Jiaotong University), Jiawei Yao (University of Washington)
TransformerLarge Language ModelVision Language ModelContrastive LearningImageTextMultimodality
🎯 What it does: By calculating the similarity between image and text embeddings, relevant image tokens are filtered to enhance the complex reasoning performance of multimodal large language models.
Enhancing Multivariate Time-Series Domain Adaptation via Contrastive Frequency Graph Discovery and Language-Guided Adversary Alignment
Haoren Guo (National University of Singapore), Tong Heng Lee (National University of Singapore)
RecognitionDomain AdaptationGraph Neural NetworkLarge Language ModelContrastive LearningTime Series
🎯 What it does: A ConFGD framework is proposed for unsupervised domain adaptation of multivariate time series, integrating techniques such as frequency graph discovery, frequency context contrastive learning, and language-guided adversarial alignment.
Enhancing NLU in Large Language Models Using Adversarial Noisy Instruction Tuning
Shengyuan Bai (Dalian University of Technology), Nianmin Yao (Hong Kong University of Science and Technology)
ClassificationRecognitionAdversarial AttackTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: A method based on Adversarial Noise Instruction Tuning (ANIT) is proposed to enhance the robustness and accuracy of large language models in natural language understanding tasks using noise instructions.
Enhancing Noise-Robust Losses for Large-Scale Noisy Data Learning
Max Staats (ScaDS.AI Dresden Leipzig), Bernd Rosenow (Institut für Theoretische Physik, Universität Leipzig)
ClassificationData-Centric LearningConvolutional Neural NetworkImage
🎯 What it does: This study investigates the learning dynamics of noise-robust loss functions in large-scale multi-class data and proposes adding a logit bias on the logits of the correct class or automatically adjusting hyperparameters to eliminate gradient vanishing and improve training effectiveness.
Enhancing Non-English Capabilities of English-Centric Large Language Models Through Deep Supervision Fine-Tuning
Wenshuai Huo (Harbin Institution of Technology), Bing Qin (Harbin Institution of Technology)
GenerationData-Centric LearningTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: A Deep Supervision Fine-Tuning (DFT) method is proposed, which explicitly adds supervision to the model's intermediate layers during non-English input, constraining both the language conversion layer (which converts the target language to English) and the English reasoning layer (which performs reasoning in the English space), thereby enhancing the non-English capabilities of English-dominant LLMs.
Enhancing Online Reinforcement Learning with Meta-Learned Objective from Offline Data
Shilong Deng (University of Electronic Science and Technology of China), Jie Shao (Duke Kunshan University)
Robotic IntelligenceMeta LearningReinforcement Learning
🎯 What it does: A GILD module is proposed, which adaptively learns a general imitation learning objective through meta-learning using offline demonstration data, thereby improving the online training effectiveness of offline RL in sparse reward environments.
Enhancing Portuguese Variety Identification with Cross-Domain Approaches
Hugo Sousa (University of Porto), Alipio Jorge (University of Porto)
ClassificationDomain AdaptationTransformerSupervised Fine-TuningText
🎯 What it does: A cross-domain Portuguese variant identification dataset PtBrVarId was constructed, and based on this, the BERTimbau fine-tuning model was used to distinguish between European Portuguese and Brazilian Portuguese.
Enhancing Question Generation through Diversity-Seeking Reinforcement Learning with Bilevel Policy Decomposition
Tianyu Ren (Queen's University Belfast), Karen Rafferty (Queen's University Belfast)
GenerationTransformerLarge Language ModelReinforcement LearningText
🎯 What it does: The BPD-DSRL framework is proposed, utilizing a dual-layer strategy decomposition and a reinforcement learning objective for diversity seeking to enhance the sample efficiency and diversity of question generation.
Enhancing Relation Extraction via Supervised Rationale Verification and Feedback
Yongqi Li (Wuhan University), Tieyun Qian (Wuhan University)
Data-Centric LearningTransformerLarge Language ModelContrastive LearningText
🎯 What it does: An automatic feedback framework SRVF based on large language models (LLM) is proposed, which can correct biased predictions in relation extraction (RE) by validating the reasoning process and providing examples for re-selection.
Enhancing Robustness in Incremental Learning with Adversarial Training
Seungju Cho (Korea Advanced Institute of Science and Technology), Changick Kim (Korea Advanced Institute of Science and Technology)
ClassificationKnowledge DistillationAdversarial AttackConvolutional Neural NetworkGenerative Adversarial NetworkImage
🎯 What it does: Proposed the ARCIL task and designed the FLAIR method to achieve adversarial robustness during the incremental learning process.
Enhancing Sequential Recommendation with Global Diffusion
Mingxuan Luo (Xiamen University), Chen Lin (Xiamen University)
Recommendation SystemTransformerDiffusion modelSequential
🎯 What it does: This paper proposes GlobalDiff, a Plug-and-Play framework that overlays diffusion models on existing sequential recommendation models. It compensates for local sequential information by restoring global non-sequential data structures, thereby improving the prediction accuracy of the next item.
Enhancing SQL Query Generation with Neurosymbolic Reasoning
Henrijs Princis (University of Cambridge), Alan Mycroft (University of Cambridge)
GenerationAI Code AssistantTransformerLarge Language ModelText
🎯 What it does: A neurosymbolic architecture called Xander is proposed, which combines symbolic reasoning and pre-trained language models to achieve SQL generation, supporting multi-path exploration, backtracking, and query repair.
Enhancing the Adversarial Robustness via Manifold Projection
Zhiting Li (Southwestern University of Finance and Economics), Guisong Liu (Southwestern University of Finance and Economics)
Knowledge DistillationAdversarial AttackAuto EncoderImage
🎯 What it does: This paper proposes the incorporation of an autoencoder for manifold projection in adversarial training and adversarial distillation to enhance model robustness.
Enhancing Trustworthiness of Graph Neural Networks with Rank-Based Conformal Training
Ting Wang (City University of Hong Kong), Rui Luo (City University of Hong Kong)
ClassificationComputational EfficiencyGraph Neural NetworkGraph
🎯 What it does: A trainable ranking-based adaptive conformal prediction framework RCP-GNN is proposed for the graph node classification task to achieve controllable improvements in boundary coverage and prediction set efficiency.
Enhancing Uncertainty Modeling with Semantic Graph for Hallucination Detection
Kedi Chen (East China Normal University), Zheng Feng
Graph Neural NetworkTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: This paper proposes a hallucination detection method based on semantic graph-enhanced uncertainty modeling. By constructing entity and sentence-level AMR semantic graphs, it improves hallucination detection performance at the sentence and paragraph levels through relation propagation and graph calibration.
Enriching Multimodal Sentiment Analysis Through Textual Emotional Descriptions of Visual-Audio Content
Sheng Wu (Tianjin University), Jianwu Dang (Shenzhen Institute of Advanced Technology)
ClassificationRecognitionTransformerLarge Language ModelTextMultimodalityAudio
🎯 What it does: Proposes the DEVA framework, which enhances audio and visual features using textual emotional descriptions, and achieves multimodal emotion analysis through text-guided progressive fusion.
Entire-Space Variational Information Exploitation for Post-Click Conversion Rate Prediction
Ke Fei (University of Electronic Science and Technology of China), Jingjing Li (University of Electronic Science and Technology of China)
Recommendation SystemKnowledge DistillationReinforcement LearningTabular
🎯 What it does: A framework for variational information utilization based on the entire space (EVI) is proposed to enhance CVR prediction accuracy through unbiased pseudo-labels and variational information maximization.
Entropy Regularized Task Representation Learning for Offline Meta-Reinforcement Learning
Mohammadreza Nakhaeinezhadfard (Aalto University), Joni Pajarinen (Aalto University)
Meta LearningReinforcement LearningGenerative Adversarial Network
🎯 What it does: This paper proposes an entropy regularization task representation learning method called ER-TRL, which utilizes GAN to approximate the entropy of the meta-behavior policy to reduce the context distribution shift in offline meta reinforcement learning, thereby enhancing adaptability and generalization ability on new tasks.
Envisioning Class Entity Reasoning by Large Language Models for Few-shot Learning
Mushui Liu (Zhejiang University), Xi Li (Zhejiang University)
ClassificationRecognitionDomain AdaptationMeta LearningTransformerLarge Language ModelContrastive LearningImageMultimodality
🎯 What it does: This paper proposes a few-shot learning framework ECER-FSL that combines specific category entities generated by large language models with abstract semantics. It significantly enhances the expressive capability of visual prototypes through layer-wise semantic-guided visual pattern extraction (SVPE) and prototype calibration (PC) modules.
EOV-Seg: Efficient Open-Vocabulary Panoptic Segmentation
Hongwei Niu (Xiamen University), Shengchuan Zhang (Contemporary Amperex Technology Co., Limited)
SegmentationComputational EfficiencyTransformerVision Language ModelImage
🎯 What it does: Proposes a single-stage shared efficient spatial awareness framework EOV-Seg to address the computational overhead and speed bottleneck of open vocabulary panoptic segmentation.
EPERM: An Evidence Path Enhanced Reasoning Model for Knowledge Graph Question and Answering
Xiao Long (University of Science and Technology of China), Shafei Wang (Peng Cheng Laboratory)
RetrievalGraph Neural NetworkTransformerLarge Language ModelSupervised Fine-TuningPrompt EngineeringTextGraphChain-of-Thought
🎯 What it does: Proposes the Evidence Path Enhanced Reasoning Model (EPERM), which divides knowledge graph question answering into three stages: retrieving and constructing subgraphs, generating and scoring evidence paths, and reasoning answers based on weighted paths.
Epistemic Bellman Operators
Pascal R. van der Vaart (Delft University of Technology), Neil Yorke-Smith (Delft University of Technology)
Reinforcement Learning
🎯 What it does: This paper proposes Epistemic Bellman Operators (EBO) to unify and theorize various Bayesian-based uncertainty RL algorithms, proving that they are contraction mappings and converge; it uses this framework to analyze and improve Bayesian Q-learning and designs an uncertainty-aware variant of PPO (ECPPO).
Epistemic EFX Allocations Exist for Monotone Valuations
Hannaneh Akrami (Max Planck Institute for Informatics), Nidhi Rathi (Max Planck Institute for Informatics)
Optimization
🎯 What it does: Proves the existence of Epistemic EFX (EEFX) allocations under any number of agents and any monotonic evaluation function, and provides a recursive construction algorithm; also proves that the EEFX allocation problem with the same submodular evaluation has an exponential query lower bound and is PLS-hard.
Epsilon: Exploring Comprehensive Visual-Semantic Projection for Multi-Label Zero-Shot Learning
Ziming Liu (Hong Kong Polytechnic University), Xiaocheng Lu (Hong Kong University of Science and Technology)
ClassificationRecognitionTransformerImage
🎯 What it does: A multi-label zero-shot learning framework named Epsilon is proposed, which combines two main modules: local semantic aggregation and global semantic diversification, to achieve a complete projection from images to semantic space.
EPT: Efficient Prompt Tuning by Multi-Space Projection and Prompt Fusion
Pengxiang Lan (Northeastern University), Xingwei Wang (Northeastern University)
TransformerSupervised Fine-TuningPrompt EngineeringText
🎯 What it does: This paper proposes an efficient prompt tuning method called EPT, which splits soft prompts into short prompts and low-rank matrices, combining prompt fusion and multi-space projection to enhance accuracy and efficiency.
Equal Merit Does Not Imply Equality: Discrimination at Equilibrium in a Hiring Market with Symmetric Agents
Serafina Kamp (University of Michigan), Benjamin Fish (University of Michigan)
Reinforcement Learning
🎯 What it does: A resource-symmetric recruitment market game model is proposed, proving that even when candidates and companies are equally strong, the game equilibrium can still produce wage discrimination.
Equilibria of the Colonel Blotto Games with Costs
Stanisław Kaźmierowski (University of Warsaw)
Optimization
🎯 What it does: This paper studies and solves the Colonel Blotto game with resource acquisition and allocation costs, proving its equivalence to a zero-sum game with an additional battlefield, thus allowing the computation of Nash equilibrium using linear programming in polynomial time.
Equirectangular Point Reconstruction for Domain Adaptive Multimodal 3D Object Detection in Adverse Weather Conditions
Jae Hyun Yoon (Chonnam National University), Seok Bong Yoo (Chonnam National University)
Object DetectionDomain AdaptationAutonomous DrivingMultimodalityPoint Cloud
🎯 What it does: This paper proposes EquiDetect, a framework for 3D object detection using LiDAR-camera multimodal fusion under adverse weather conditions.
Erase Then Rectify: A Training-Free Parameter Editing Approach for Cost-Effective Graph Unlearning
Zhe-Rui Yang (Sun Yat-sen University), Hao Liu (Hong Kong University of Science and Technology)
OptimizationComputational EfficiencyGraph Neural NetworkGraph
🎯 What it does: A two-stage graph unlearning method called ETR is proposed, which first eliminates the target samples and their propagation effects through parameter editing, and then corrects the model performance using subgraph gradient approximation.
ERF: A Benchmark Dataset for Robust Semantic Segmentation Under Extreme Rainfall Conditions
Xin Yang (National University of Singapore), Xinchao Wang (National University of Singapore)
SegmentationDomain AdaptationTransformerImageVideoBenchmark
🎯 What it does: This paper proposes and constructs the first semantic segmentation dataset ERF under real heavy rain scenarios, and systematically evaluates the robustness of multi-class image and video models in this extreme environment.
ERL-MPP: Evolutionary Reinforcement Learning with Multi-head Puzzle Perception for Solving Large-scale Jigsaw Puzzles of Eroded Gaps
Xingke Song (University of Nottingham Ningbo China), Xudong Jiang (Nanyang Technological University)
OptimizationTransformerReinforcement LearningGenerative Adversarial NetworkImage
🎯 What it does: This paper proposes a framework based on Evolutionary Reinforcement Learning and Multi-Head Puzzle Perception (ERL-MPP) to address the large-scale puzzle reconstruction problem with significant gaps.
Error Analysis Affected by Heavy-Tailed Gradients for Non-Convex Pairwise Stochastic Gradient Descent
Jun Chen (Huazhong Agricultural University), Weifu Li (Huazhong Agricultural University)
Optimization
🎯 What it does: Analyzed the generalization error and optimization error of pairwise stochastic gradient descent (SGD) with heavy-tailed gradient noise (sub-Weibull) in non-convex dual learning, and provided corresponding theoretical upper bounds.
Error Bounds for Gaussian Process Regression Under Bounded Support Noise with Applications to Safety Certification
Robert Reed (University of Colorado Boulder), Morteza Lahijanian (Delft University of Technology)
OptimizationSafty and PrivacyTabular
🎯 What it does: This paper studies the error bounds of Gaussian Process Regression (GPR) under bounded support noise, deriving new probabilistic and deterministic error bounds, and applying these bounds for safety probability assessment in safety control systems.
Error Diversity Matters: An Error-Resistant Ensemble Method for Unsupervised Dependency Parsing
Behzad Shayegh (University of Alberta), Lili Mou (University of Alberta)
Text
🎯 What it does: A multi-model ensemble for unsupervised dependency parsing has been constructed, and a social entropy-based error diversity-driven ensemble selection method has been proposed.
ESEG: Event-Based Segmentation Boosted by Explicit Edge-Semantic Guidance
Yucheng Zhao (Beijing University of Technology), Yongjian Deng (Southeast University)
SegmentationConvolutional Neural NetworkTransformerImage
🎯 What it does: This paper proposes the ESEG framework, which utilizes the motion edge characteristics of event cameras, introduces explicit semantic edge supervision, and enhances event-driven semantic segmentation performance through a multi-layer fusion module.
ESPRESSO: An Effective Approach to Passage Retrieval for High-Quality Conversational Recommender Systems
Taeho Kim (Hanyang University), Sang-Wook Kim (Hanyang University)
RetrievalRecommendation SystemTransformerContrastive LearningTextRetrieval-Augmented Generation
🎯 What it does: This paper presents ESPRESSO, a retrieval module specifically designed for conversational recommendation systems (CRS), aimed at enhancing the authenticity and information richness of recommendation responses by retrieving paragraphs that match user preferences.
Evaluating Image Hallucination in Text-to-Image Generation with Question-Answering
Youngsun Lim (Kim Jaechul Graduate School of AI KAIST), Hyunjung Shim (Kim Jaechul Graduate School of AI KAIST)
GenerationData SynthesisTransformerLarge Language ModelVision Language ModelImageTextBenchmark
🎯 What it does: This paper proposes an evaluation framework called I-HallA to assess whether image generation models produce factual errors through visual question answering, and constructs the I-HallA v1.0 benchmark dataset.
Evaluating LLM Reasoning in the Operations Research Domain with ORQA
Mahdi Mostajabdaveh (Huawei Technologies Canada), Yong Zhang (Huawei Technologies Canada)
OptimizationTransformerLarge Language ModelPrompt EngineeringTextBenchmarkChain-of-Thought
🎯 What it does: This paper evaluates the reasoning and knowledge transfer capabilities of large language models (LLMs) in the field of operations research (OR) and proposes a specialized multiple-choice question-answering benchmark—ORQA, which is used to systematically assess various open-source LLMs.